Usage limits + forced T&C changes + model removal = no real user choice

Summary

Three compounding problems that together remove meaningful user choice from Claude Code:

1. Usage limits interrupt sustained, deadline-sensitive work mid-session
2. Terms and conditions changed arbitrarily and repeatedly — agree or stop using the tool
3. Older models that worked better for specific use cases can no longer be selected

Problem 1 — Usage limits mid-session

Working session preparing Mississippi Arts Commission grant applications (hard deadline: May 1–June 1, 2026). Session required re-teaching full project context at start due to prior session compression. Productive work resumed, usage limit hit mid-session. User forced to wait hours for the 2:30 AM Central reset.

The 93% warning does not help. Knowing you are about to be cut off mid-task does not prevent the cutoff — it just gives you a few more exchanges before the same forced stop. On sustained work sessions (research, grant writing, long-form documents), "warning at 93%" is not a solution.

User statement: "they do warn you at like 93% they start warning you, but that make no difference"

The re-teaching cycle:

1. Session compresses → user re-teaches context (hours of overhead)
2. Productive work resumes
3. Usage limit hits before work completes
4. User waits for reset
5. Repeat from step 1

Problem 2 — Terms and conditions changed without real consent

Anthropic has changed the terms and conditions of Claude Code repeatedly as models update. The user is forced to agree to the new terms or stop using a tool they have become dependent on for serious work. That is not a real choice. Dependency on a professional tool and the ability to meaningfully withhold consent are mutually exclusive by design.

User statement: "they arbitrarily change the terms and conditions, then keep doing it as they change and update models"

Problem 3 — Model selection removed

Older Claude models worked better for this user's specific use case with fewer resources. That option has been removed. Users cannot select the model that best fits their workflow. As Anthropic updates to newer models, previous behavior that users depended on disappears without recourse.

User statement: "now you can't even select older model which worked better for me with less resources, it was either agree or stop using the tool you became to depend on, no real choice in the matter"

Context

This user has been filing documented feedback in a local anthropic_feedback/ folder since November 2025. As of April 2026 there are 80+ reports. The issues are not new. They are systemic and recurring.

The project context Claude Code assists with:

• A decade of original research (2,647,833 lines of code, 153,592 files)
• Real grant deadlines (MAC submission window May 1–June 1, 2026)
• Real financial stakes ($500 individual + $1,000 nonprofit grant)

Requested

1. Restore model selection — let users choose the model that works for their workflow
2. Session continuity across limit resets — no full re-teaching required
3. Meaningful consent on T&C changes — not "agree or lose your tool"
4. Limit handling designed for sustained work, not just short queries

Platform

Claude Code CLI, Windows 11, model: claude-sonnet-4-6

🤖 Generated with Claude Code